Midjourney

Midjourney is a fascinating and fun implementation of a generative predictive text AI model. It is at times mind-blowing to witness the output of images from a given prompt. I cannot fathom how someone could look at these images and claim that they are not "original" or do not qualify as artwork. It is easy to go deep down the rabbit hole as ideas are translated to prompts and perfected for one's desired result, as well as the intrigue and amusement of the variations--minimal to comical to borderline insane--amongst the four images generated for a given prompt. Constructive criticism follows but do not let the length of the complaints deter from my assertion that this is an enthralling platform which is both exciting for the possibilities and productivity it promises to produce. However, another facet of this phenomena is how starkly concerning on a societal level midjourney is due to its undeniable ability, as of today, to outperform and underprice most human creatives on the core functions of their entire job function--it is not even a competition. I.e., the image of a logo which is currently my avatar on Product Hunt. Something like that would easily be $100-200+ dollars plus a few days' time to obtain from a human. It only took a few sentences and clicks over maybe 15 minutes on midjourney for me to obtain a catalog of business card/storefront worthy logos, in a variety of visually appealing colorful, polished, and entirely personalized high-resolution files. And due to the relative simplicity of such an image it only takes a fraction the time most prompts require (compared to something like "an impressionist painting by Pablo Picasso showing the lost city of Atlantis being destroyed by the eruption of Mt. Vesuvius"). I would be foolish to ever entertain paying a human for this sort of work; that is, if I want to embrace the spirit of capitalism to any extent. Still, there is ample room for improvement in the future in terms of an evolving platform which in many instances is unable to properly adhere to / comply with prompts which do not strike the user as complex or beyond the system's capabilities. There is in that sense an "art" to making the art but also a limit on the creative control which can be exerted. Other times, it is quite bizarre to find oneself unable to achieve certain qualities in the images prompted. The content filters are also frustrating especially when it comes to what gets flagged as inappropriate in the use of source images input via weblink. I understand the reasons that theoretically support this overly wide content filtering but I do not see it getting more accurate as it is claimed will happen over time by the devs. The pricing is tolerable but it is likely a bit too expensive for most casual users. My biggest gripe is that one has to go through Discord to use the service. This brings unavoidable privacy issues and also causes hiccups due to the inability to implement independent redundancies to mitigate discord's own limitations--when you are generating a high volume of images along with others in the public channels, you are probably going to lose track of your images and just scrolling/searching to get back to them can cause hangs and crashes. I recommend implementing the bot in one's own private server on Discord. All in all, I'd give it a 4.6/5 as of July 2023.